Journal of Digital Information Management Michael Granitzer, Andreas S. Rath, Mark Kröll, Christin Seifert, Doris Ipsmiller, Didier Devaurs, Nicolas Weber, Stefanie Lindstaedt 7 5 2009 http://www.dirf.org/jdim/v3n4a2.asp Increasing the productivity of a knowledge worker via intelligent applications requires the identification of a user’s current work task, i.e. the current work context a user resides in. In this work we present and evaluate machine learning based work task detection methods. By viewing a work task as sequence of digital interaction patterns of mouse clicks and key strokes, we present (i) a methodology for recording those user interactions and (ii) an in-depth analysis of supervised classification models for classifying work tasks in two different scenarios: a task centric scenario and a user centric scenario. We analyze different supervised classification models, feature types and feature selection methods on a laboratory as well as a real world data set. Results show satisfiable accuracy and high user acceptance by using relatively simple types of features.